Modeling Atmospheric O2 Over Phanerozoic Time

Home , Robert Berner

Geochimica et Cosmochimica Acta, Vol. 65, No. 5, pp. 685–694, 2001 Copyright © 2001 Elsevier Science Ltd Pergamon Printed in the USA. All rights reserved 0016-7037/01 $20.00 ϩ .00 PII S0016-7037(00)00572-X

Modeling atmospheric O2 over Phanerozoic time

R. A. BERNER* Department of Geology and Geophysics, Yale University, New Haven, CT 06520-8109, USA

(Received May 18, 2000; accepted in revised form October 9, 2000)

Abstract—A carbon and sulfur isotope mass balance model has been constructed for calculating the variation

of atmospheric O2 over Phanerozoic time. In order to obtain realistic O2 levels, rapid sediment recycling and O2-dependent isotope fractionation have been employed by the modelling. The dependence of isotope fractionation on O2 is based, for carbon, on the results of laboratory photosynthesis experiments and, for sulfur, on the observed relation between oxidation/reduction recycling and S-isotope fractionation during early diagenetic pyrite formation. The range of fractionations used in the modeling agree with measurements of Phanerozoic sediments by others.

Results, derived from extensive sensitivity analysis, suggest that there was a positive excursion of O2 to levels as high as 35% during the Permo-Carboniferous. High O2 at this time agrees with independent modeling, based on the abundances of organic matter and pyrite in sediments, and with the occurrence of giant insects during this period. The cause of the excursion is believed to be the rise of vascular land plants and the

1. INTRODUCTION 2000). Most of this work has focused on the control of organic production in the oceans by nutrient elements such as phos- The evolution of atmospheric oxygen over geologic time has phorus or nitrogen. However, in these studies calculations of been a subject of major interest to both the earth and biologic O levels as a function of time were not constrained by estimi- sciences. This is because O is both a principal cause and 2 2 nation of actual fluxes of both carbon and sulfur as they affect principal effect of biologic evolution. On shorter time scales the input and output of O to and from the atmosphere. atmospheric oxygen concentration represents the balance be- 2 The calculation of past atmospheric oxygen concentrations is tween production by photosynthesis and consumption by res- possible using the approach pioneered by Garrels and Lerman piration. On a long geologic time scale the level of O is 2 (1984). This method uses the isotopic record of carbon and affected by both geological and biologic processes. It is pro- sulfur in sedimentary rocks to calculate the rates of the produced by net photosynthesis (photosynthesis minus respira- cesses that affect O on long time scales, mainly burial in tion), represented by the burial of organic matter and pyrite 2 marine sediments and weathering on land of organic matter and (FeS ) in sediments, and it is consumed by the oxidation of old 2 biogenic pyrite. Using the Garrels and Lerman model, esti- buried sedimentary organic matter and pyrite upon uplift and mates of the rates of weathering and burial, and their effect on exposure to weathering on the continents. The oxidation of O , have been made by Berner (1987), Francois and Gerard reduced gases produced by the thermal decomposition of or- 2 (1986), Kump and Garrels (1986), Kump (1989), Lasaga ganic matter and pyrite at great depths also exerts a control on (1989) and Petsch and Berner (1998). However, none of these O . (The formation and reduction of iron oxides have only a 2 studies has presented a realistic history of O variations over minor effect on O —Holland, 1978). 2 2 Phanerozoic time. This is because a key difficulty in this type Although much attention has been given to the rise of O and 2 of modelling is formulating negative feedback mechanisms life during the Precambrian (e.g., Berkner and Marshall, 1965; consistent with both geologic and biologic processes and the Cloud, 1976; Walker, 1977; Schidlowski, 1984; Holland, 1984; isotopic record of carbon and sulfur. Without feedback the Des Marais et al.,1992; Canfield and Teske, 1996), little atten- modelling leads to calculated variations in O over Phanerozoic tion has been paid to its continued evolution during the Pha- 2 time that are biologically and physically impossible, such as nerozoic. Some previous studies (Berner and Canfield, 1989) large negative values for O concentration during the early have suggested that atmospheric oxygen concentrations may 2 Paleozoic. have changed appreciably over the Phanerozoic whereas others Initial attempts to include negative feedback in carbon-sulfur (e.g., Watson et al., 1978; Lenton and Watson, 2000) have models have let the rate of organic matter and pyrite weathering suggested that O has been held essentially constant by strong 2 be a direct function of the concentration of atmospheric O .In negative feedbacks. 2 other words as O rises, the rate of its consumption by weath- Recently there has been a lot of interest in negative feedback 2 ering also rises. However, this approach, although it brings mechanisms and how they might have controlled Phanerozoic about somewhat damped variations in O , still results in im- O levels (Holland, 1994; Van Cappellen and Ingall, 1997; 2 2 possible O histories (Lasaga, 1989; Berner and Petsch, 1998). Falkowski, 1997; Colman et al., 1998; Lenton and Watson, 2 With O2-dependent weathering, positive feedback is unavoid- ably introduced because of the need for carbon and sulfur *Author to whom correspondence should be addressed isotope mass balance. (For example, if organic C burial sud- ([email protected]). denly increased, atmospheric O2 would increase and organic C 685 686 R. A. Berner

weathering would also increase if O2-dependent weathering (Canﬁeld and Teske, 1996) that, in the presence of O2, sedi- were present. Although this would help to lower atmospheric mentary sulfur undergoes increased isotope fractionation be-

O2, it would also lead to an increased input to the ocean of light cause of multiple cycles of sulfate reduction and sulﬁde oxi- carbon from the extra organic weathering necessitating, for dation accompanying the overturn of sediments near the steady state, futher increased organic C burial and O2 produc- sediment/seawater interface due to bioturbation and wave and tion—in other words, positive feedback). In addition, doubt has current stirring. As atmospheric O2 increases, it is assumed been placed on whether the rate of O2 uptake by weathering is that, on a global average basis, the extent of oxidation/reduc- actually related to atmospheric O2 level (Holland, 1978; Hol- tion cycling increases which results in an increase in overall land, 1984; Chang and Berner, 1999). Unavoidable positive isotope fractionation between seawater sulfate and the pyritic feedback is also introduced if the rate of burial of organic sulfur ultimately buried in sediments.

matter and/or pyrite are assumed to be an inverse function of From the studies of processes controlling O2, it is reasonable atmospheric O2. to assume that calculated past O2 levels should lie within One way to add unencumbered negative feedback to Garrels certain extremes. In the present paper limits are defined by the and Lerman type isotope modelling is to introduce rapid recy- response of forest fires to changing O2. At sufficiently low O2 cling (Berner, 1987). In rapid recycling (see also, Berner and concentration forest fires will not take place, but there is evi- Canfield, 1989) it is assumed that younger rocks that have been dence of fires, in the form of fossil charcoal (Chaloner, 1989), deposited more recently are weathered preferentially. In this for all periods since the rise of woody plants about 425 million way rapid variations in the isotopic composition of sediments years ago. On the other hand, O2 levels could not have been so as they are buried do not result in equally large complementary high that the destruction of all terrestrial life by fires took place variations in oceanic isotopic composition because the just- (Watson et al., 1978; Kump, 1989; Robinson, 1989). From deposited rocks are soon weathered and the anomalous isotopic these studies and our earlier modeling study (Berner and Can- composition rapidly returned to the ocean. However, calcula- field, 1989) we have adopted the range 10 to 40% O2 as a tions show that this approach alone is insufficient to produce reasonable outside limit for the testing of our model calcula- realistic O2 levels and something more is needed. tions. A mathematical adjustment to Garrels and Lerman-type In the present paper calculations of the concentration of O2 modelling, that mimics negative feedback while conserving over Phanerozoic time are done using the Garrels and Lerman

carbon and sulfur mass balance, is to let the fractionation of model, modifed to accommodate rapid recycling and O2-de- isotopes during organic matter and pyrite formation be a func- pendent C and S isotope fractionation. By sensitivity analysis,

tion of the level of atmospheric O2. This idea is reasonable. the relative importance of each modiﬁcation is illustrated. From 12 Plant biomass is enriched in C relative to CO2 due to signif- this analysis, curves of O2 vs. time can be obtained that are in icant isotope discrimination during photosynthesis (e.g., Far- agreement with laboratory measurements of isotope fraction- quhar et al., 1982). If rates of plant respiration (photorespira- ation and that agree with independent calculations based on the

tion) increase in response to elevated O2/CO2, then less rubisco abundance of organic carbon and pyrite sulfur in sedimentary is available for CO2 ﬁxation and as a result CO2 builds up in the rocks (Berner and Canﬁeld, 1989). cell and carbon isotope discrimination is increased. Also, with continued photosynthesis, recycling of the extra photorespira- 2. METHOD OF CALCULATION

tion-derived CO2 may result in plant biomass that is more enriched in 12C than would be predicted based only on the The method of Garrels and Lerman (1984) as adapted by Berner (1987) was used to calculate the rates of weathering and sediment isotopic composition of ambient, extracellular CO2. Using this burial of organic carbon and pyrite sulfur (see also Berner and Petsch, reasoning, we have studied the effect of changing O2 on carbon 1998). From these data the value of the mass of atmospheric oxygen isotope discrimination and found enhanced fractionation for (O2) as a function of time can be calculated from the expression: both vascular land plants and a marine diatom at elevated O 2 d ͑O ͒/dt ϭ Fbg Ϫ Fwg ϩ ͑15/8͒͑Fbp Ϫ Fwp͒ (1) levels (Berner et al., 2000). This has allowed us to introduce a 2 dependence on atmospheric O2 level to Garrels and Lerman where: type modelling. Fbg ϭ burial rate of organic C in sediments Since the response of fractionation to photosynthesis and Fwg ϭ rate of weathering of organic C (plus minor oxidation of photorespiration is a function of both O2 and CO2, fractionation reduced C-containing gases released by volcanism, metamor- can also change with changing CO2 levels (Hayes et al., 1999). phism, and deep diagenesis) However, the purpose of this paper is to focus on O , and as Fbp ϭ burial rate of pyrite-s in sediments 2 Fwp ϭ rate of weathering of pyrite S (plus minor oxidation of reduced will be seen below, the largest change in fractionation over the S-containing gases released by volcanism, metamorphism, and Phanerozoic, that occurring during the Permo-Carboniferous, is deep diagenesis) best explained in terms of changes in O2, not CO2. (This The concentration of O is calculated from its mass in the atmosphere conclusion has been veriﬁed by unpublished preliminary cal- 2 by assuming a constant mass of N over the Phanerozoic. From the very culations that let carbon isotope fractionation be a function of 2 long residence time of N2 in the atmosphere (about 300 million years) O2 and CO2 using new laboratory measurements on plants, the relative to uptake and release to/from rocks (Holland, 1978) this as- ␦13 determination of C of plant fossils, and the Phanerozoic CO2 sumption is reasonable. data of Berner, 1997). The Garrels and Lerman method uses mass balance expresssions for total carbon and sulfur input to the oceans, and output to sediments, and One can also add O2-dependence to the fractionation of the 13C/12C and 34S/32S composition of these inputs and outputs. sulfur isotopes during bacterial sulfate reduction and pyrite Calculations are based on the isotopic composition of carbon and sulfur formation (Berner et al., 2000). This is based on the observation in the oceans as a function of time as recorded by carbonates and Modelling Panerozoic atmospheric O2 687

Fig. 1. Simple ﬁrst order ﬁt to the ␦13C measurements of Veizer et al. (1999) on unrecrystallized carbonate fossils. The curve is used in the modeling of the present paper. (Plotted data after Veizer et al., 1999).

sulfates in sedimentary rocks. In the present study isotopic data for the tive feedback to the system by allowing for the fact that, due to sea Phanerozoic were taken for carbon from the results of Veizer et al. level changes, sediments deposited more recently are more likely to be (1999) for the Ordovician-Cretaceous (Fig. 1) and Lindh (1983) for the exposed to weathering on the continents since they overly the older remainder of the Phanerozoic. The carbon isotopic data are greatly sediments. smoothed to eliminate shorter term variations and, thereby, to show An important additional modification to the Garrels and Lerman only the first order Phanerozoic trend. (For the Ordovician-Cretaceous modelling was done by making the fractionation of carbon istopes only the data actually measured by Veizer et al. are used here, and not between carbonate and organic matter (due to photosynthesis) and those summarized by them from the literature, because of the greater sulfur isotopes between sulfate and sulfide (due to bacterial sulfate care taken by Veizer et al. in selecting samples for isotopic analysis). reduction) be a function of the level of atmospheric O2. (The effect of Sulfur isotope data are from the compilation by Lindh (1983). CO2 on fractionation is discussed below.) Here is employed a function In Garrels and Lerman modeling carbon is added to the ocean/atm that agrees with the results of plant and plankton growth experiments system by the weathering of carbonates and organic carbon exposed on (Berner et al., 2000) and photosynthetic theory for land plants (Far- land combined with degassing of C-containing gases via the metamor- quhar et al., 1982; Farquhar and Wong, 1984; see also Berner et al., phic, volcanic and diagenetic breakdown of deeply buried organic C 2000): and carbonates. Carbon is removed by the burial of organic matter and ⌬ Ϸ ␦13 Ϫ ␦13 ϭ ⌬ ͑ ͒ Ca and Mg carbonates in sediments. Sulfur is added to the ocean/atm c ͑phot͒ C CO2 Corganic c͑phot͒ 0 system by the weathering of pyrite (FeS2) and CaSO4 minerals (plus ϩ ͑͑ ͒ Ϫ ͒ degassing from the thermal breakdown of both minerals) and removed J O2/38 1 (2) by the burial of newly formed mixtures of these minerals in sediments. where: At each time step the masses of the four reservoirs, organic carbon, carbonate carbon, sulfide sulfur, and sulfate sulfur, and their isotopic ⌬ ϭ c(phot) isotope fractionation (discrimination) of carbon (in ‰) dur- compositions are recalculated. From these data, combined weathering ⌬ ϭ⌬ ing photosynthesis c(phot)(0) c (phot) value for the plus metamorphic release rates are calculated from the assumption of a present level of O2 but for an average higher CO2 level direct proportionality of flux to the mass of each reservoir present at during the Phanerozoic (Berner, 1997; Hayes et al., 1999) each time. The calculations are set up so that present day masses and ϭ O2 mass of oxygen in the atmosphere fluxes are obtained at the end of each computer run. 38 ϭ mass of oxygen in the present atmosphere (in 1018 moles) Negative feedback was added to the Garrels and Lerman model by J ϭ adjustable curve-fit parameter employing rapid recycling (Berner, 1987; Berner and Canfield, 1989). In this method the mass of each reservoir is divided into young, rapidly Values of J were varied until there was agreement of the expression weathering and old, slowly weathering subreservoirs. The young sub- with the results of the plant and plankton growth experiments. The reservoirs were assumed to be 25% the mass of the old reservoirs and physical interpretation of J is complex; for present purposes it can be 13 to weather ten times faster. This means that the mean residence time of thought of simply as the strength of the effect of O2 on C isotope the fast-weathering reservoirs to loss by weathering was assumed to be fractionation during photosynthesis for a fixed level of CO2. (Presently 35 to 75 million years. This is a reasonable representation of the undertaken laboratory experiments on plants by D. J. Beerling show area-weighted mean age of sedimentary rocks presently exposed to that J can be expressed as a function of CO2, without causing major weathering at the earth’s surface. All burial of carbon and sulfur change in the calculated results for O2 over time.) necessarily goes to the young reservoirs. This method provides nega- Because the modeling uses the difference in ␦13C between calcium 688 R. A. Berner

Fig. 2. Sensitivity of O2 vs. time to the addition of rapid recycling and O2-dependence of carbon isotope fractionation ϭ ϭ ⌬ ϭ 18 (J 3, n 1) to isotope mass balance modeling with c(0) 25‰. O2 is expressed in 10 moles. The situation for no ϭ ϭ O2-dependence is signiﬁed by J 0, n 0. For the deﬁnitions of J and n see text Eqns. 2 and 3.

⌬ carbonate and organic matter buried in sediments ( c), it is necessary and current stirring is enhanced by greater levels of atmospheric O2 ⌬ ⌬ to calculate the value of c from c(phot). This involves correcting for (Canfield and Teske, 1996), then a reasonable first approximation is to the fractionation between CO2 and CaCO3 which varies with temper- let fractionation be a direct function of O2 concentration. This is what ature, and for changes in ␦13C of organic matter due to food-chain is done here. processing and diagenesis (Hayes et al., 1999). Because this paper As a check on our modeling of carbon and sulfur isotope fraction- focuses only on O2-dependent photosynthetic fractionation, these other ations we have used compilations for the measured difference in fractionations are assumed to remain constant over time. (A complete isotopic composition between organic carbon and carbonate carbon model for the change of fractionation with time would allow for (Hayes et al., 1999) and pyrite sulfur and CaSO4-sulfur (Strauss, 1999) temperature change, changes in organic processing, and changes in for Phanerozoic sediments. This “ground truth” provides an additional ⌬ CO2 as discussed below.) On this basis, c can be substituted for test to the results of our modeling. ⌬ ⌬ ⌬ c(phot) and c(0) for c(phot)(0) in Eqn. 2 because of the assumption of a constant offset over time between ⌬ (phot) and ⌬ . c c 3. RESULTS For sulfur a simple positive dependence function was used, based on the observation of greater fractionation with higher atmospheric O 2 Figures 2–9 show the sensitivity of varying different input (Canfield and Teske, 1996) parameters to the values of O2 mass as a function of time. In ⌬ Ϸ ␦34 Ϫ ␦34 ϭ ⌬ ͑ ͒n s Ssulfate Ssulfide s͑0͒ O2/38 (3) Figures 2 to 4 oxygen is expressed as total mass in the atmo- 18 where: sphere in 10 moles. Figure 2 shows the effects of adding rapid recycling and oxygen dependency to both carbon and sulfur ⌬ ϭ s fractionation for sulfur (in ‰) ⌬ ϭ isotope fractionation. The symbols J and n represent the curve s(0) fractionation for the present level of O2 (35‰) n ϭ arbitrary parameter fit parameters of Eqn. 2 and (3). Figures 3 is an expansion of Figure 2 for later times and also includes the effects of O2- Experiments to derive a better defined dependence would be virtually dependent C and S isotope fractionation alone. Figures 4 and 5 impossible because of the occurrence under natural conditions of (1) the variability of isotope fractionation during sulfate reduction which illustrate the variation of carbon isotope fractionation over time depends on such things as rate of reduction, nature of the carbon source based on varying values of J in Eqn. 2 and with comparsion to and temperature (Canfield et al., 2000); (2) the extent of sulfate reduc- the C isotope data of Hayes et al. (1999). Figures 6 and 7 show tion (closed vs open system in sediments); (3) the time of pyrite how sulfur isotope fractionation varies with varying values of n formation (syngenetic vs diagenetic); and (4) additional fractionation accompanying sedimentary reoxidation/disproportionation during the in Eqn. 3 with comparison of results to the results of Strauss recycling of sulfur in sediments by bioturbation and wave and current (1999). In Figure 8 sensitivity to change in the effect of the stirring. If reoxidation/disproportionation from bioturbation and wave value chosen for the fractionation of carbon isotopes at present Modelling Panerozoic atmospheric O2 689

ϭ ϭ Fig. 3. Sensitivity of O2 vs. time plots to the addition of O2-dependence of carbon isotope fractionation (J 3, n 0), ϭ ϭ ϭ ϭ ⌬ ϭ sulfur isotope fractionation (J 0, n 1) and both (J 3, n 1) for the period of 400 my to the present. ( c(0) 25‰, rapid recycling).

⌬ ϭ ϭ ⌬ ϭ Fig. 4. Plots of carbon isotope fractionation c vs. time for J 3, n 1 and c(0) 25‰, 30‰ compared with the measured values for sedimentary carbonates and bulk organic matter compiled by Hayes et al. (1999). 690 R. A. Berner

⌬ ϭ ⌬ ϭ Fig. 5. Plots of carbon isotope fractionation c vs. time for different values of J and n 1. ( c(0) 25‰, rapid recycling).

⌬ c(0) is demonstrated. Finally best estimates of atmospheric O2 shown by the Hayes et al. data is attributed mainly to the effect over Phanerozoic time, in terms of % O2, are shown in Figure of higher CO2 levels over most of the geologic past in produc- 9 and compared to the results of Berner and Canﬁeld (1989) ing enhanced carbon isotope fractionation during photosynthe- based on sediment abundance data for organic carbon and sis (Freeman and Hayes, 1992). All three curves show a posi- pyrite sulfur. tive excursion during the middle Paleozoic centered around 300

Ma which we believe was due to higher levels of O2 at that 4. DISCUSSION time. Using constant values over time for the fractionation of It is less likely that the 300 ma excursion is due to a rise in carbon and sulfur isotopes, the effect of including rapid recy- CO2 which could also bring about greater fractionation. This is cling in carbon and sulfur isotope mass balance modeling partly because terrestrially-derived plant debris was the major source for buried organic matter during this time (Berner and results in an improved curve of O2 vs. time, but the results still shows impossibly large excursions (Fig. 2). The excursions are Raiswell, 1983) and land plants exhibit much lower sensitivity greatly reduced by adding the dependence of carbon and sulfur of carbon isotope fractionation to changes in atmospheric CO2 ϭ ϭ (Polley et al., 1993; Beerling. 1996; Arens et al., 2000). Also, isotope fractionation on O2 (J 3 and n 1—see Eqn. 2 and 3). From Figure 3 it is apparent that the improvement in the independent studies (for a summary consult Berner, 1997) indicate that there was most likely lower CO during this period O2-vs time curve is due to the inclusion in the modelling of 2 oxygen-dependent isotope fractionation for both carbon and compared to CO2 levels previously and subsequently. Further, sulfur but with a greater effect for sulfur for the early Paleozoic. unpublished experiments of carbon isotope fractionation, as a Carbon isotope fractionation vs time, expressed as carbon- function of O2/CO2, by a variety of land plants combined with ate-C minus organic-C, calculated using J ϭ 3 in Eqn. 2 and determination of the ␦13C of fossil plants (Beerling, ms in n ϭ 1 in Eqn. 3 is shown in Figure 4 and compared with the preparation) show that variations in O2, and not CO2 best data of Hayes et al. (1999). For sensitivity analysis two values explain the inctrease in fractionation during the Permo-Carbon- ⌬ of c(0) were used: 25‰ and 30‰. So as to retain the simple iferous. It is important to note that the use of either the Hayes O2-dependence of our model, no attempt was made to match et al. curve or either of ours from Figure 4 results in similar ⌬ the actual changes of c reported by Hayes et al., especially plots of O2 vs. time (see below). ⌬ those over the Cenozoic. The generally higher past values Figure 5 demonstrates how the variation of c with O2 over Modelling Panerozoic atmospheric O2 691

⌬ ϭ ⌬ ϭ Fig. 6. Plots of sulfur isotope fractionation s vs. time for different values of n and J 3. ( c(0) 25‰, rapid recycling).

time depends on the value of J in Eqn. 2. Laboratory measure- variation of O2 over Phanerozoic time based on Garrels and ments of carbon isotope discrimination by vascular land plants Lerman isotope mass balance modeling with rapid recycling.

(one flowering plant and one cycad) and marine plankton (a Also included in Figure 9 is a plot of O2 vs. time from the species of diatom) are fit best by J ϭ 2 to 3 (Berner et al., results of Berner and Canfield (1989). The Berner and Canfield 2000). Because of this, the curve labelled J ϭ 3 in Figure 5 is results are based on a completely independent method where preferred. burial fluxes are calculated from the abundance of organic C Sensitivity of sulfur isotope fractionation to variations in the and pyrite S in sedimentary rocks as a function of time, and no exponent n is shown in Figure 6. The variation of the measured isotopic data are used. One goal of the present study was to see degree of sulfur isotope fractionation between sedimentary if a plot of O2 vs time could be derived from isotope mass pyrite and seawater sulfate (as recorded by sulfate in evaporites balance modelling which could be compared with that based on and incorporated into calcium carbonate) has been summarized rock abundance. By choosing the appropriate expressions for by Strauss (1999) for Phanerozoic sediments. In Figure 7 the O2 dependence of C and S isotope fractionation and em- Strauss’ results are compared to the calculated value for J ϭ 3, ploying rapid recycling, it was possible to obain three curves in n ϭ 0.5. Considering the many factors that affect sulfur isotope good agreement, with the rock abundance-based results (within fractionation during sedimentary pyrite formation and how they the errors of the Berner and Canfield estimates). This is shown could have varied over Phanerozoic time (Canfield and Teske, in Figure 9 and it gives some credence to the idea that both 1996), the calculated and observed data fall well within the methods of theoretical modeling provide some idea of the same range. (Note the very large range in the results compiled variation of atmospheric O2 over Phanerozoic time. by Strauss for each time period which would likely expand as The most obvious feature of Figure 9 is the large maximum more pyrite samples in older rocks are analyzed.) in O2 centered around 300 million years BP. The explanation Sensitivity of O2 vs time to change in the present carbon for this has been given earlier (Holland, 1978; Berner, 1987; isotope fractionation between carbonate-C and organic-C Berner and Canfield, 1989) and involves the rise of large ⌬ ϭ ( c(0) 25‰ and 30‰) is shown in Figure 8. The similar vascular land plants on the continents. The production of abun- results demonstrate the relative insensitivity to the value chosen dant new sources of microbially resistant organic matter, espe- for the Phanerozoic mean value (see Fig. 4) which, in each case cially lignin, should have led to greater burial of organic C, is similar to that for the present day. This is agrement with the both in terrestrial swamps and in the sea after being carried earlier sensitivity studies of Kump (1989) and Lasaga (1989). there by rivers. This greatly enhanced burial led to the forma- ϭ ϭ ⌬ ϭ Figure 9 shows three best estimates (J 3, n 0.5, c(0) tion of vast coal deposits (the largest for all geologic time) and ϭ ϭ ⌬ ϭ ⌬ ϭ 25‰; J 3, n 1, c(0) 30‰; and Hayes c,n 1) of the the excessive production of O2. The evidence for this is the 692 R. A. Berner

⌬ ϭ ϭ Fig. 7. Plot of sulfur isotope fractionation s vs. time for J 3, n 0.5 compared with the compilation of sulfur isotope measurements by Strauss (1999).

⌬ ϭ ϭ ϭ Fig. 8. Plots of O2 vs. time for c(0) 25‰ and 30‰, J 3 (Eqn. 2) and n 1 (Eqn. 3). This shows the relative insensitivity of O2 to the present value for carbon isotope fractionation (see Fig. 4). Modelling Panerozoic atmospheric O2 693

Fig. 9. Plots of % O2 vs. time for the Phanerozoic calculated in three ways by isotope mass balance modeling (present study) and by the abundance of organic C and pyrite S in sedimentary rocks (Berner and Canfield, 1989 denoted as ⌬ ⌬ Canfield). J, n, and c(0) refer to Eqn. 2 and 3 of the text; Hayes refers to c values reported by Hayes et al. (1999). (Constant mass of atmospheric N2 is assumed in calculating %O2 from its atmospheric mass.). large increase in 13C of the ocean (plus atmosphere) at that time ated respiration, and the invasion of land by vertebrates at this 12 (Berner, 1987) due to an increased removal of C from sea- time may have been due to elevated O2. In agreement with this water and the atmosphere as organic matter buried in sedi- reasoning, Harrison and Leighton (1998) have found from ments. This idea is confirmed by the calculations of Berner and experiments that elevated O2 enhances flight metabolism in Canfield (1989) who show that, due the abundance of organic dragonflies and they have used this observation as a possible carbon-rich coal basin sediments, global organic C burial was explanation for the presence of giant dragonflies (up to 70 cm considerably elevated during the Permo-Carboniferous. This wing span) that flourished during the Carboniferous. In addition helps explain the agreement of our O2 curve with that of Berner to animal physiologic considerations, Beerling et al. (1998) and Canfield. have shown that plants can survive and grow at elevated O2 Such high levels of O2 for the Permo-Carboniferous have concentrations as suggested here for the Permo-Carboniferous. encountered objections. Watson et al. (1978) and Lenton and Watson (2000) have stated that levels of O2 could not have 5. CONCLUSIONS risen at any time higher than about 25% because otherwise vast global forest fires would have wiped out all plant life. Their Isotope mass balance calculations along with the earlier results are based on the experimental ignition of paper strips at calculations based on the abundances of pyrite and organic

different O2 levels and degrees of moisture. However, the matter in sedimentary rocks, have led to the conclusion that the application of paper strip ignition to the ecology of forest level of atmospheric O2 has varied considerably over Phanero- burning has been severely criticized by Robinson (1989) who, zoic time, reaching a peak during the Permo-Carboniferous that on the contrary, has presented evidence for more ﬁre-resistant was brought about by the rise of large vascular land plants. The plants during the Permo-Carboniferous as evidence of higher calculated concentrations fall within reasonable limits, but

O2 levels. We feel that experiments on burning under more much more modeling is needed before one can say anything natural conditions are needed before one can place firm limits more definite about absolute past concentrations of O2. Better on forest fire frequency as a function of O2 level. models are needed that include the effects of varying CO2 and Some recent physiologic studies have shown that higher O2 temperature on isotope fractionation (Hayes et al., 1999) and levels during the Permo-Carboniferous could help explain a the effects of factors, such as climate and relief, that affect number of observations of fossil animals. Graham et al. (1995) weathering rates (e.g., Berner, 1994). Also needed to guide the and Dudley (1998) have suggested that much of the gigantism modeling is more experimental plant and plankton growth of insects, changes in marine organisms with diffusion-medi- studies and more experimental animal metabolism studies, both 694 R. A. Berner conducted on organisms representative of those living during cellular carbon-dioxide concentration in leaves. Aust. J. Plant. the distant geological past. These studies. along with further Physiol. 9, 121–137. investigation of the response of forest fires to changing O Farquhar G. D. and Wong S. C. (1984) An empirical-model of stomatal 2 conductance. Aust. J. Plant. Physiol. 11, 191–209. levels, should go far in helping to further knowledge of the Francois L. M. and Gerard J. C. (1986) A numerical model of the evolution of the atmosphere over geologic time. evolution of ocean sulfate and sedimentary sulfur during the past 800 million years. Geochim. Cosmochim. Acta 50, 2289–2302. Acknowledgments—I acknowledge Steven Petsch, John Hayes, Dave Freeman K. H. and Hayes J. M. (1992) Fractionation of carbon isotopes DesMarais, Kate Freeman and an anonymous reviewer for discussions by phytoplankton and estimates of ancient CO2 levels. Glob. Bio- and helpful reviews of the manuscript. Research was supported by geochem. Cyc. 6, 185–198. DOE Grant DE-FG02 95ER14522 and NSF Grant EAR 9804781. Garrels R. M. and Lerman A. (1984) Coupling of the sedimentary sulfur and carbon cycles-an improved model. Am. J. Sci. 284, 989– 1007. Associate editor: K. Freeman Graham J. B., Dudley R., Aguilar N. M., and Gans C. (1995) Impli- cations of the late Paleozoic oxygen pulse for physiology and evo- REFERENCES lution. Nature 375, 117–120. Harrison J. F. and Lighton J. R. B. (1998) Oxygen-sensitive flight Arens N. C., Jahren A. H., and Amundson R. (2000) Can C3 plants metabolism in the dragonfly Erythemus simplicicollis. J. Exp. Biol. faithfully record the carbon isotopic composition of atmospheric 201, 1739–1744. carbon dioxide? Paleobiology 26, 137–164. 13 Hayes J. M, Strauss H., and Kaufman A. J. (1999) The abundance of Beerling D. J. (1996) C discrimination by fossil leaves during the 13 late-glacial climate oscillation 12–10 ka BP: Measurements and C in marine organic matter and isotope fractionation in the global physiological controls. Oecologia. 108, 29–37. biogeochemical cycle of carbon during the past 800 Ma. Chem. Geol. Beerling D. J., Woodward F. I., Lomas M. R., Wills M. A., Quick 161, 103–125. W. P., and Valdes P. J. (1998) The influence of Carboniferous Holland H. D. (1978) The chemistry of the atmosphere and oceans. paleoatmospheres on plant function: An experimental and modeling New York: Wiley Interscience. assessment. Phil. Trans. Roy. Soc. Series B 353, 131–140. Holland H. D. (1984) The chemical evolution of the atmosphere and Berkner L. V. and Marshall L. C. (1965) On the origin and rise of oceans. Princeton, Princeton Univ. Press, 582 p. oxygen concentration in the Earth’s atmosphere. J. Atm. Sci. 22, Holland H. D. (1994) The phosphate-oxygen connection. EOS, Trans. 225–261. Am. Geophys. Union. 75, No.3., p. OS96 Berner R. A. (1987) Models for carbon and sulfur cycles and atmo- Kump L. R. (1988) Terrestrial feedback in atmospheric oxygen regu- spheric oxygen: Application to Paleozoic geologic history. Am. J. lation by fire and phosphorus. Nature 335, 152–154. Sci. 287, 177–190. Kump L. R. (1989) Alternative modeling approaches to the geochemi- Berner R. A. (1997) The rise of plants and their effect on weathering cal cycles of carbon, sulfur and strontium isotopes. Am. J. Sci. 289, 390–410. and atmospheric CO2. Science 276, 544–546. Berner R. A. and Canfield D. E. (1989) A new model of atmospheric Kump L. R. and Garrels R. M. (1986) Modeling atmospheric O2 in the oxygen over Phanerozoic time. Am. J. Sci. 289, 333–361. global sedimentary redox cycle. Am. J. Sci. 286, 336–360. Berner R. A. and Petsch S. T. (1998) The sulfur cycle and atmospheric Lasaga A. C. (1989) A new approach to the isotopic modeling of the oxygen. Science 282, 1426–1427. variation of atmospheric oxygen through the Phanerozoic. Am. J. Sci. Berner R. A., Petsch S. T., Lake J. A., Beerling D. J., Popp B. N., Lane 298, 411–435. R. S., Laws E. A., Westley M. B., Cassar N., Woodward F. I., and Lenton T. M. and Watson A. J. (2000) Redfield revisited 2. What Quick W. P. (2000). Isotope fractionation and atmospheric oxygen: regulates the oxygen content of the atmosphere? Glob. Biogeochem. Implications for Phanerozoic O evolution. Science 287, 1630–1633. Cycles. 14, 249–268. 2 13 34 Berner R. A. and Raiswell R. (1983) Burial of organic carbon and Lindh T. B. (1983) Temporal variations in C and S and global pyrite sulfur in sediments over Phanerozoic time: a new theory. sedimentation during the Phanerozoic. Thesis. Univ. Miami. 98 pp Geochim. Cosmochim. Acta 47, 855–862. (unpublished). Canfield D. E., Habicht K. S., and Thamdrup B. (2000) The. Archean Petsch S. T. and Berner R. A. (1998) Coupling the long-term geo- sulfur cycle and the early history of atmospheric oxygen. Science chemical cycles of carbon, phosphorus, sulfur, and iron: the effect on 288, 658–661. atmospheric O2 and the isotopic records of carbon and sulfur. Am. J. Canfield D. E. and Teske A. (1996) Late Proterozoic rise in atmo- Sci. 298, 246–262. spheric oxygen concentration inferred from phylogenetic and sul- Polley H. W., Johnson H. B., Marino B. D., and Mayeux H. S. (1993) phur-isotope studies. Science 382, 127–132. Increase in C3 plant water-use efficiency and biomass over glacial to Chaloner W. G. (1989) Fossil charcoal as an indicator of paleoatmo- present CO2 concentrations. Nature. 361, 61–64. spheric oxygen level. J. Geol. Soc. London 146, 171–174. Robinson J. M. (1989) Phanerozoic O2 variation, fire, and terrestrial Chang S. and Berner R. A. (1999) Coal weathering and the geochemi- ecology. Paleogeog. Paleoclim. Paleoecol. 75, 223–240. cal carbon cycle. Geochim. Cosmochim. Acta 63, 3301–3310. Schidlowski M. (1984) Early atmospheric oxygen levels: constraints Cloud P. E. (1976) Beginnings of biospheric evolution and their bio- from Archaen photoautotrophy. J. Geol. Soc. London. 141, 243–250. geochemical consequences. Paleobiology 2, 351–387. Strauss H. (1999) Geological evolution from isotope proxy signals— Colman A. S., Mackenzie F. T., and Holland H. D. (1998) Redox sulfur. Chem. Geol. 161, 89–101. stabilization of the atmosphere and oceans and marine productivity. Van Cappellen P. and Ingall E. D. (1997) Redox stabilization of the Science 275, 406–407. atmosphere and oceans by phosphorus-limited marine productivity. Des Marais D. J., Strauss H., Summons R. E., and Hayes J. M. (1992) Science 271, 493–496. Carbon isotope evidence for the stepwise oxidation of the Protero- Veizer J., Ala D. Azmy K., Bruckschen P., Buhl D., Bruhn F., Carden zoic environment. Nature 359, 605–609. G. A. F., Diener A., Ebneth S., Godderis Y., Jasper T., Korte C., Dudley R. (1998) Atmospheric oxygen, giant Paleozoic insects and the Pawellek F., Podlaha O. G., and Strauss H. (1999) 87Sr/86Sr, ␦13C evolution of aerial locomotor performance. J. Exp. Biol. 201, 1043– and ␦18O evolution of Phanerozoic seawater. Chem. Geol. 161, 1050. 59–88. Falkowski P. G. (1997) Evolution of the nitrogen cycle and its influ- Walker J. C. G. (1977) Evolution of the Atmosphere. New York,

ence on the biological sequestration of CO2 in the ocean. Nature 387, Macmillan. 272–275. Watson A., Lovelock J. E., and Margulis L. (1978) Methanogenesis, Farquhar G. D., O’Leary M. H., and Berry J. A. (1982) On the ﬁres and the regulation of atmospheric oxygen. Bio. Systems. 10, relationship between carbon isotope discrimination and the inter- 293–298.